Google has recently published an article explaining how their popular "Busyness Indicator" works. For instance, Google maps location information as well as search queries are mentioned as relevant data sources. Based on that article, I can roughly imagine how such data can be used to estimate historical footfall, residence or waiting times.

What is entirely unclear to me is how predictions are made, or at least how temporal patterns are generated. I suspect that opening times, day of the week, season (summer vs. winter), holidays (e.g., Christmas), but also prevailing weather conditions have an impact on footfall.

Is there any evidence on which factors are considered (or generally should be for high prediction accuracy), either among the ones I've mentioned before or not, and do we know anything about models that would be appropriate for such a purpose? For instance, I could imagine some kind of Fourier regression that considers the impact of weekday/time as well as seasonalities, with additional constants for specific holidays or dependence on temperature/precipitation.

I'm mentioning Google's solution in particular as I'm a frequent user of it myself. Having said that, I'd be just as interested in the statistical models used by any other, similar solution. Many thanks.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.